LETTER doi:10.1038/nature11205 Reconciling the temperature dependence of respiration across timescales and ecosystem types Gabriel Yvon-Durocher 1,2 *, Jane M. Caffrey 3 , Alessandro Cescatti 4 , Matteo Dossena 1 , Paul del Giorgio 5 , Josep M. Gasol 6 , Jose ´ M. Montoya 6 , Jukka Pumpanen 7 , Peter A. Staehr 8 , Mark Trimmer 1 , Guy Woodward 1 & Andrew P. Allen 9 * Ecosystem respiration is the biotic conversion of organic carbon to carbon dioxide by all of the organisms in an ecosystem, including both consumers and primary producers. Respiration exhibits an exponential temperature dependence at the subcellular and individual levels 1 , but at the ecosystem level respiration can be modified by many variables 2–4 including community abundance and biomass 5 , which vary substantially among ecosystems 6 . Despite its importance for predicting the responses of the bio- sphere to climate change, it is as yet unknown whether the temper- ature dependence of ecosystem respiration varies systematically between aquatic and terrestrial environments. Here we use the largest database of respiratory measurements yet compiled to show that the sensitivity of ecosystem respiration to seasonal changes in temperature is remarkably similar for diverse environments encompassing lakes, rivers, estuaries, the open ocean and forested and non-forested terrestrial ecosystems, with an average activation energy similar to that of the respiratory complex 3 (approximately 0.65 electronvolts (eV)). By contrast, annual ecosystem respiration shows a substantially greater temperature dependence across aquatic (approximately 0.65 eV) versus terrestrial ecosystems (approximately 0.32 eV) that span broad geographic gradients in temperature. Using a model 5 derived from metabolic theory 7 , these findings can be reconciled by similarities in the biochemical kinetics of metabolism at the subcellular level, and fundamental differences in the import- ance of other variables besides temperature—such as primary pro- ductivity and allochthonous carbon inputs—on the structure of aquatic and terrestrial biota at the community level. We assessed variability in the temperature dependence of ecosystem respiration within and among a range of aquatic and terrestrial environ- ments using a global compilation of measurements of respiration from nine distinct ecosystem types that represent entire ecosystems or eco- system components (Supplementary Information 1). We performed two analyses using these data. First, we assessed the sensitivity of eco- system respiration to seasonal changes in temperature within sites for each ecosystem type, and quantified its variation among sites and across ecosystem types using daily (hereafter short-term) estimates of flux. Second, we determined the temperature sensitivity of respira- tion at longer timescales by comparing annual (hereafter long-term) fluxes across sites spanning broad geographic gradients in temper- ature. We reconcile the similarities and differences in the temperature dependence of ecosystem respiration across timescales and ecosystem types using a model 5 derived from metabolic theory 7 . To determine variation among sites in the seasonal temperature dependence of ecosystem respiration for the nine ecosystem types in our compilation, we fit the short-term respiration data to the Boltzmann–Arrhenius function using linear mixed-effects modelling 8 (see Methods Summary, Supplementary Information 2 and 5): ln R s (T )~(  E R ze s E )(1=kT C {1=kT )z ln R(T C )ze s R ð1Þ In this expression, lnR s (T) is the natural logarithm of respiration rate for some arbitrary site s at absolute temperature T (in kelvin (K)),  E R is an average among sites for the apparent activation energy, which characterizes the temperature sensitivity of ecosystem respiration, and k is the Boltzmann constant (8.62 3 10 25 eV K 21 ). We centred the temperature data using a fixed, arbitrary value (5 288 K 5 15 uC) so that ln R(T C ) corresponds to an average among sites for the rate of ecosystem respiration at 15 uC, R(T C ). We would expect R(T C ) to vary among sites due to factors that affect the availability of reduced carbon substrates to support biomass, including net primary production 9,10 and allochthonous carbon inputs 6,11–13 , as well as factors that affect the susceptibility of reduced carbon substrates to decomposition by biota, such as C:N:P stoichiometry 6 and water availability 14 . We would also expect R(T C ) to vary seasonally within a site 2,3 , resulting in a deviation of the apparent activation energy from  E R (Supplementary Information 2), owing to processes that co-vary with temperature, such as litterfall and nutrient turnover in the water column 15 . To account for these factors in our linear mixed-effects models, we treated the slope and intercept as random variables with averages of  E R and ln R(T C ), respectively, and site-specific deviations from these averages of e s E and e s R for each site s. Analyses of the short-term data revealed marked similarities in the seasonal temperature dependence of ecosystem respiration across all nine ecosystem types (Fig. 1). Estimates of the average apparent activa- tion energy,  E R , were statistically indistinguishable from each other (likelihood ratio test; x 2 8 5 7.36, P 5 0.50), with an average of 0.62 eV (Table 1), which corresponds to a Q 10 —that is, the proportional increase in respiration per 10 uC rise in temperature—of ,2.5 at 15 uC. Consistent with our model, the apparent activation energy varied between sites, as reflected by the significance of the term used to represent e s E in eight of the nine models (Table 1), but this variation was not systematically different among ecosystem types (Supplemen- tary Information 8 and 9). Recent work indicates that this variability partly reflects localized factors—for example, water availability, pro- ductivity, allochthonous carbon input—that seasonally co-vary with respiration, R(T C ), and temperature, and can modulate the apparent temperature sensitivity at the site level 2,3 . Our model yields predictions on how the magnitude of this covariation affects the apparent activa- tion energy at a given site, and thus provides a biological interpretation for differences among sites (Supplementary Information 2). To set our results in a more general theoretical context, we can explore them further by applying a model derived from metabolic theory 5 . Because metabolic theory relates complex ecosystem-level phe- nomena to the effects of body mass and temperature on individual-level *These authors contributed equally to this work. 1 School of Biological & Chemical Sciences, Queen Mary University of London, London E1 4NS, UK. 2 Environment and Sustainability Institute, University of Exeter, Penryn, Cornwall TR10 9EZ. UK. 3 Center for Environmental Diagnostics and Bioremediation, University of West Florida, 11000 University Parkway, Pensacola, Florida 32514, USA. 4 European Commission, Joint Research Centre, Institute for Environment and Sustainability, Ispra I-21027, Italy. 5 De ´ partement des sciences biologiques, Universite ´ du Que ´ bec a ` Montre ´ al, Montre ´ al, Province of Que ´ bec, H2X 3X8, Canada. 6 Institute of Marine Sciences (ICM-CSIC), Pg. Marı ´tim de la Barceloneta, 37-49 E-08003 Barcelona, Spain. 7 University of Helsinki Department of Forest Sciences, PO Box 27, FI-00014 University of Helsinki, Finland. 8 Aarhus University, Institute of Bioscience, Frederiksborgvej 399, PO Box 358, 4000 Roskilde, Denmark. 9 Department of Biological Sciences, Macquarie University, Sydney, New South Wales 2109, Australia. 00 MONTH 2012 | VOL 000 | NATURE | 1 Macmillan Publishers Limited. All rights reserved ©2012